One spring morning, a dozen students gathered around tables in a shared classroom, their eyes glued to math lessons on their laptops.
Sixth graders at Khan Lab School, an independent school with an elementary campus in Palo Alto, California, were working on quadratic equations, graphing functions and Venn diagrams. But when faced with questions, many did not immediately call their teacher for help.
They used a text box next to their lessons to request help from Khanmigo, an experimental chatbot tutor for schools using artificial intelligence.
The tutor robot quickly responded to a student, Zaya, asking her to identify specific data points in a graph. Then Khanmigo persuaded her to use the data points to solve her math question.
“It’s very effective in walking you step by step through the problem,” Zaya said. “Then he praises you every time he helps you solve a problem.”
Khan Lab School students are among the first schoolchildren in the United States to try experimental conversational chatbots aimed at simulating one-on-one human tutoring. The tools can respond to students in clear, fluent sentences, and they have been designed specifically for school use.
Based on AI models underlying chatbots like ChatGPT, these automated study aids could usher in a profound change in teaching and learning in the classroom. Simulated tutors could allow many self-directed students to hone their skills, delve deeper into topics that interest them, or tackle new topics at their own pace.
Such unproven, automated tutoring systems could also make mistakes, encourage cheating, diminish the role of teachers, or hinder critical thinking in schools, forcing students to test out topics for what amounts to a learning experience. education by algorithm. Or, like a legion of promising tech tools before them, robots might simply not do much to improve educational outcomes.
Khanmigo is part of the wave of new AI-based learning tools. It was developed by Khan Academy, a nonprofit education giant whose video tutorials and practice problems have been used by tens of millions of students.
Sal Khan, founder of Khan Academy — and Khan Lab School, a separate nonprofit — said he hopes the chatbot will democratize student access to one-on-one tutoring. He also said it could greatly help teachers with tasks such as lesson planning, allowing them to spend more time with their students.
“This will allow every student in the United States, and eventually on the planet, to benefit from a world-class personal tutor,” Mr. Khan said.
Hundreds of public schools already use Khan Academy’s online courses for math and other subjects. Today, the nonprofit, which introduced Khanmigo this year, is testing the tutoring robot with districts, including public schools in Newark, New Jersey.
Khan Academy developed the robot with guardrails for schools, Mr Khan said. These include a monitoring system designed to alert teachers if students using Khanmigo seem obsessed with issues such as self-harm. Mr. Khan said his group was studying Khanmigo’s effectiveness and planned to make it widely available in districts this fall.
Thousands of U.S. schools already use analytical AI tools such as plagiarism detection systems and adaptive learning applications designed to automatically adjust lessons to students’ reading levels. But advocates see new AI-assisted tutoring systems as a game-changer in education because they act more like student collaborators than inert software.
AI’s facility with language has prompted some enthusiasts to say that simulated tutors could soon be as responsive to students as human tutors.
“AIs will achieve this ability, to be as good a tutor as any human being,” said Bill Gates, Microsoft co-founder and philanthropist. said at a recent conference for investors in educational technologies. (Khan Academy has received more than $10 million in grants from the Bill & Melinda Gates Foundation.)
It’s not yet clear whether robots can provide the kind of empathetic support and genuine encouragement that can make human tutors particularly effective.
For more than a century, education entrepreneurs have imagined classroom devices programmed to automatically test students and deliver instruction.
As educational writer Audrey Watters recounts in her book “Teaching machines“, researchers in the 1920s began claiming that automated teaching devices would revolutionize education. The machines, they promised, would free teachers from drudgery and allow students to work at their own pace and receive automated comments.
Over the decades, schools that have rushed to adopt the latest automated teaching technologies have often found that capricious or faulty systems. Some concluded that automated tools did little to improve student outcomes.
Today, new chatbots are sparking a new push for automated educational tools. Khanmigo highlights the educational promise and potential downsides of technology.
Khan Academy began developing chatbot tutoring software last fall with the goal of evaluating the potential of AI to improve learning. The system uses GPT-4, a large language model created by OpenAI, the research lab behind ChatGPT.
Mr Khan said he wanted to create a system to help guide students, rather than just giving them answers. So the developers at Khan Academy designed Khanmigo to use the Socratic method. He often asks students to explain their thinking to encourage them to solve their own questions.
Khanmigo offers help on a wide variety of topics: elementary school math, middle school American history, high school civics, and college organic chemistry. It also has features that invite students to chat with fictional characters like Winnie the Pooh or simulated historical figures like Marie Curie.
AI systems based on large linguistic models can also concoct false information. This is because the models are designed to predict the next word in a sequence. They don’t stick to the facts.
To improve Khanmigo’s math accuracy, Khan Academy developers created a multi-step process: the system develops answers to a math problem behind the scenes, then compares them to a student’s answer. Despite this, Khan Academy’s tutoring system displays a warning at the bottom of the screen: “Khanmigo sometimes makes mistakes. »
Khan Lab School, where annual tuition costs more than $30,000, provides an ideal testbed for tutoring robots. The Silicon Valley school offers small class sizes and an entrepreneurial philosophy encouraging children to pursue their passions and learn at their own pace. Its tech-savvy students are used to tinkering with digital tools.
One spring morning, Jaclyn MajorSTEM specialist at Khan Primary School, watched as her students playfully tested the robot’s limits.
One student asked Khanmigo to explain a math problem using lyrics from a song. Another asked for help with math in “Gen Z slang.”
“Will you do me an extra favor and explain everything in Korean?” said a third during a text conversation with the chatbot.
Khanmigo dutifully obliged. Then it brought each student back to the math task at hand.
Ms Major said she liked how the system interacted in an engaging way with her students.
“Khanmigo is able to connect with them and be on their level if they want to be,” she said. “I think this could be useful in any classroom.”
It’s too early to tell whether Khanmigo will be equally appealing to other audiences, such as public schools with larger class sizes or students who are not used to driving their own learning.
In the class, Zaya, a sixth grader, had encountered a problem. Khanmigo had asked her to explain how she found the answer to a dataset problem. Then the robot incorrectly suggested that she may have made a “small error” in her calculations.
She quickly chastised the AI chatbot: “19 + 12 is 31 khanmigo,” she wrote.
“Sorry for my mistake earlier,” Khanmigo replied. “You are indeed right.”
This could prove to be one of the most important lessons for schoolchildren using promising new robot tutors: don’t believe every AI-generated text you read.
“Remember, we’re testing it,” Ms. Major reminded her students. “We are learning – and it is learning.”